A WEB-BASED APROACH FOR ONLINE DIGITAL TERRAIN MODEL AND
ORTHOIMAGE GENERATION
Z. Paszotta, M. Szumilo *
Dept. of Photogrammetry and Remote Sensing, University of Warmia and Mazury in Olsztyn,
ul. Oczapowskiego 1,10-719 Olsztyn, paszotta@uwm.edu.pl, malgorzata.szumilo@uwm.edu.pl
Commission IV, WG IV/5
KEY WORDS: Internet/Web, Matching, DEM/DTM, Orthoimage, Programming, Orthorectification
ABSTRACT:
Currently, the Digital Terrain Model (DTM) and orthoimages have become standard geospatial products. However,
photogrammetric systems used for generating them still remain desktop or workstation based systems. This paper outlines an
innovative concept of DTM generation and geospatial image orthorectification over the Internet. The article contains description of
algorithms for automatic DTM and orthoimage construction. DTM is an element that is necessary in the process of creating
orthoimages and has a crucial influence on their cartometric attributes. Thus, it is extremely important to determine sources of errors
that can influence the accuracy of the Digital Terrain Model constructing photogrammetric methods. Last part of the article contains
presentation of Internet application dedicated to DTM and orthoimage construction and their quality assessment. Several Java clientserver web applications were developed. In practice the application of the proposed algorithms is based on the system of a central
unit (a server of applications and Web pages) and customer computers equipped with standard Internet browsers. It is possible to
access this application from the Internet site of Department of Photogrammetry and Remote Sensing UWM in Olsztyn
(http://www.kfit.uwm.edu.pl/zp.html) by selecting the “Orthophoto generation” projects. Opportunities of networked DTM
generation and image ortorectification have had a great impact on the development of geo-information processing.
1. INTRODUCTION
New technologies have always been a key driving force in
geographic information science. The Internet has opened up
fantastic new opportunities for applications. There is an
abundance of powerful, inexpensive personal computers and
mobile devices that can access the Internet and run these new
applications. The Internet also provides a new possibility for
visualization of aerial and satellite images and other mapping
datasets (Dai, et al., 2007). The needs of geospatial information
obtain in the near real-time is also important for example to
change detection or environmental monitoring. One of such
data source is ortophoto image and DTM (Digital Terrain
Model) that have became standard geospatial products.
However, photogrammetric systems have used for generating
them still remain desktop or workstation based systems. People
in particular geodetic, architects or town planners, who wants to
use topical orthophoto image to his study, has neither suitable
photogrametric photos nor appropriate software to process
them. The opportunity of networked orthorectification come
with modern information technologies has had a great impact
on the development of geo-information processing.
This paper outlines an innovative concept of DTM generation
and geospatial image orthorectification over Internet. The
article contains description of algorithms for automatic DTM
and orthoimage construction and presentation of Web
application based on algorithms mentioned previously.
2. ARCHITECTURE OF WEB PHOTOGRAMMETRIC
APPLICATION
The nature of web-based applications lend themselves to an ntiered approach (Petersen, 2009). A typical application usually
includes three essential elements: presentation, logic and data.
Presentation represents user interface, logic refers to processing,
and data refers to database or database management system.
The relationship between these elements is that one element
makes a request to the other element and the other element then
fulfills the requests. The element that makes request is called a
client and the element that fulfills the request is called server.
A web application is one of the example the client/server model
implementation (Peng, Z.,Tsou, M., 2003). A 3-tier architecture
is the most common approach used for such applications today.
In the typical example of this model, the web browser acts as
the client, an application server handles the logic, and a separate
tier (database server) handles database functions (Petersen,
2009).
In the physical client/server model, a client refers to the
computer that interacts with the end user and the server refers to
the computer that manages and stores data and/or other
processing functions.
In constructing a Web photogrammetric application, it can be
assumed that the best quality of photogrammetric data
(especially aerial photos and very high resolution satellite
imageries) was stored on expensive, well-equipped data servers.
Software that is necessary to read data from these servers was
installed on another computer called an application server.
Moreover, users have been communicating with the application
server by means of their own software. Because we have a webbased application in mind, the client used a web browser for
* Corresponding author. This is useful to know for communication with the appropriate person in cases with more than one author.
this task. When the client communicated with the application
server the HTML code of the web side and short programs called applets have been sent as a response. Actions on the
server side have been conducted by programs called servlets.
Applets and servlets are written in the JAVA programming
language and compiled to bytecode, which at runtime is either
interpreted or compiled to native machine code for
execution.(Graba 2003)
The scheme of distributed application described above was
shown in the figure 1
Figure 1. Architecture of web photogrammetric application.
In order to the web application presented in this paper acted
correctly it is necessary to have a WWW page server, e.g.
Apache HTTP Server, and an application server, e.g. Apache
Tomcat to execute servlets. Additionally, users had to install
Java Virtual Machine (JVM) to execute applets on the client
side. In presented solution the Microsoft SQL Server fulfilled
the database server role.
3. DESCRIPTION OF TEST DATASET
AND ITS PROCESSING
To generate orthophoto image the following data are needed:
digital aerial photos, elements of interior and external photos
orientation and DTM. The test material included the block of
two strips photos. There were six photos in the strip. The aerial
photos were taken by an analog Zeiss’s LMK 2030 camera in
the region of campus of University of Warmia and Mazury in
Olsztyn. The scale of photos was 1:8000, with the camera focus
equaling approximately 305 mm. Coordinates of principal point
and fiduacial marks were included in the attached calibration
report of the camera. Photos were scanned with the pixel size
approximately 21 µm.
To determine the transformation from pixel coordinate system
to an image coordinate system, affine transformation was used.
The coefficients of inverse transformation were calculated too.
Additionally, for determination of the external orientation
aerotiangulation was performed.
The all data concern photos were stored in the database.
According to the structure presented in the figure database
contained two tables. Table PHOTO included basis information
like numbers of columns, rows and levels of pyramid, and also
shown which photos made up stereo-pair. There are elements of
photos
orientation
in
the
second
table,
where
Figure 2. The structure of photos database.
4. ALGORITHM OF DTM AND ORTHOIMAGE
GENERATION
DTM and orthophoto images a commonly produced by the use
of digital photogrammetric workstations that are usually
equipped with advanced technology. The process of creating
web photogrametric applications has not been based on copying
or translating the source codes of such systems. It was
necessary to build new programs to conduct definite tasks. They
ought to fill gaps in of the user’s existing software and take into
consideration the properties of the network structure (Harold
2004).
The algorithm of the web-based approach for online DTM and
orthoimage generation is presented with the aid of UML
activity diagram in the figure 3. In presented solution there are
three successive steps: parameters determination, DTM built
and orthoimage generation. In the next part of paper we focus
on the problem of DTM constraction. The mathematical basis
of construction of orthoimage as a function was also described.
a _ af , b _ af , c _ af , d _ af , e _ af , f _ af , ck , xs, ys
related to coefficients of affine transformation, whereas
X 0, Y 0, Z 0, OM , FI , KA was the external orientation
elements.
Figure 3. Activity diagram shows steps in the process of online
DTM and orthoimage generation
4.1 Digital Terrain Model
c' , c"
DTM is an element that is necessary in the process of creating
orthoimages. Aerial photogrammetry provides major source of
data that are used to create it. The main task in the
photogrammetrical approach is to measure homologous points
on two or more images (photos), i.e. selecting an object on one
photo and finding its equivalent on the other one. In analogous
and analytical photogrammetry, this is achieved by manual
measurements
performed
by
an
operator.
Digital
photogrammetry aims to solve this problem automatically. This
process is called image matching.
Algorithms for automatic DTM construction from raster data
(aerial photographs, satellite imagery) are mainly based on
correlation methods. The main measurement of similarity (goal
function), while using the image matching method, is a
coefficient of correlation. For a defined neighbourhood of one
pixel on the left photo (neighbourhood of grid point on the left
photo) and the same neighbourhood on the right photo, a
coefficient of correlation is repeatedly calculated. The area is
considered to be matched for the maximum value of the
coefficient and when its value is larger than the threshold value
(Paszotta, 2006).
The theoretical basis and detailed explanation of mentioned
image matching algorithm was presented in other article
(Paszotta 2000a)
The above described procedure is repeated on subsequent,
previously established grid nodes. Matching results in obtaining
a dataset, which includes pixel coordinates of the centres of the
areas matched on the left picture, recording the differences
between coordinates on the right and on the left picture and the
minimum of the goal function (Paszotta, 2000b)
After image coordinates of homologous points have been
obtained, data to generate DTM can be acquired in two ways:
a spatial model is built, by determining elements of
reciprocal orientation, which is then transformed to
the ground coordinate system.
after determining the elements of exterior orientation
of photos, image coordinates of points measured on
the stereogram are directly converted into X,Y,Z
coordinates in a ground system, by using collinearity
equation.
The last step of DTM generation is modellling of terrain
surface. There are many methods of DTM interpolation widely
described in the literature of the subject.
- fragments of the left and right photo of stereopair.
4.2 Orthorectification process
The orthophoto image is obtained in the orthorectification
process. We defined orhtoimage and digital photo as functions
c' (i ' , j ' ) , c(i, j ) where arguments of these functions are
indexes of pixels (pixel coordinate) and their values belong to
the colors space. The colors of orthoimage pixels c ' : (i ' , j ' ) is
received on the basis of color pixels of digital photo
This relation can be described as the function
c': (i ' , j ' ) → c(e( g ( f (i ' , j ' ))))
c : (i, j ) .
(2)
According to the function 2 the process of orthoimage
generation
can
be
shown
in
the
following
steps (Paszotta, et al. 2010). The function f presented the
coordinates of the center of the orhtoimage pixel in the ground
coordinate system.
f (i ' , j ' ) ⇒ X , Y
(3)
The next step was to determine Z ( X , Y ) based on Digital
Terrain Model.
g : ( X , Y , Z ) ⇒ ( x, y )
Figure 4. Algorithm of image matching
The diagram of image matching process is shown in the figure
4. We assumed there that the goal function has the form
d (c' c" ) = 1 − r (c' c" )
where
r (c' c" )
- correlation coefficient
(1)
(4)
The function g could be described by collinearity equation. It
allowed to transform ground coordinates of the point into its
digital photo coordinates in the fiducial marks coordinates
system.
Then based on the function e the digital photo coordinates were
transformed to digital photo pixel coordinate.
e : ( x, y ) ⇒ (i, j )
(5)
The last step was calculated the color of digital photo pixel used
one of the resampling method c : (i, j ) ⇒ c (i, j )
As a result of the ortorectification process the image that show
cartometric view of the Earth’s surface is recorded in BMP
format.
5. THE EXAMPLE OF ONLINE DTM AND
ORTHOIMAGE GENERATION
The presented web application has consisted of applets and
servlets that was written in the JAVA programming language
and has worked on the Department of Photogrammetry and
Remote Sensing serverhttp://www.kfit.uwm.edu.pl/zp1/or.html.
Data that is necessary for constructing orthoimage has been
obtained from several sources: parameters of photos
orientations from the database, photos and their fragments from
a directory of photos on the dataserver, and other data – initial
parameters of process of DTM and orthoimage generation have
been delivered by the client. The application has started
working on the client side after downloading the mentioned
above web site. User has had to select the stereopair and
appropriate photo pyramid level, indicate the area of interest on
the selected photos and send to application server request for
preparation of photos fragment. After that the web browser
windows looked like in the figure 5.
The next step is DTM generation. The main task of this stage
concentrate on homologous point measurement. The measure of
similarity between the digital images is the correlation
coefficient. To perform this operation, pixel color values from
RBG color space should be transformed to monochromatic
color space. The user defines also a threshold for correlation
coefficient, the target size (pattern), and step changes in the
coordinates (in pixels) on the right photo (size of search
window). After selecting a starting point on the left and right
photos – identifications of all other conjugate points is
performed automatically. The result of this operation is shown
in the figure 6. It should be emphasize that the image matching
is executed on the original images on the server side.
Simultaneously the terrain coordinates of points that were used
to create digital terrain model (DTM) were calculated in the
automatic way.
Figure 6. Results of image matching shown
in the web browser window
Since the area of the orthophotomap is small and limited by
dimension of the window of the web browser. Thus, in the
process of orthorectification over the Internet we assumed a
simplified DTM. The proposed method of digital terrain model
is approximated it in the form of the plane.
Z = AX + BY + C
Figure 5. View of the web browser window with the running
web application of DTM and orthoimage generation.
(6)
Figure 7. Points selection for plane approximation.
Control point for finding the plane equation coefficients are
selected by the user through assignments them the weight
equal 1 (figure 7). The coefficients of the plane equation are
defined by deploying a least squares adjustment method.
Accuracy assessment of the model is also implemented.
Conformity between the established model and remaining point
was determined by using RMS error. In presented example the
RMS is equal 0.9m. (figure 7).
The last stage of our web application is ortorectification
process.
For the orthoimage generation, user needs to
interactively setup the extent of the working area and define a
scale of the result, as it is shown in figure7. As a result of
orthorectification process the carthometric image is recorded in
BMP format. Additionally the compressed orthoimage is
produced in JPEG format and through the applet was
automatically forwarded to the client to be displayed in the web
brower window.
Because of resampling, the created
orthoimage has a worse quality than a source photo. Therefore,
the source photograph with the system of coordinates is also
presented.
7. REFERENCES
Dai, N., Mandel, L., Ryman, A., 2007 Eclipse Web Tool
Platform. Developing JAVA Web Applications. Addison
Wesley, Stoughton, Massachusetts, USA
Graba, J., 2003. An introduction to network programming with
Java, Addison Wesley, ISBN 0 321 11614 3
Harold, E.R., 2004. Java Network Programming, 3rd Edition.
O’Reilly Media Inc. ISBN 978-0596007205
Paszotta, Z., 2000a , Teoretyczne podstawy metody spasowania
obszarów obrazów cyfrowych.
Archiwum Fotogrametrii,
Kartografii i Teledetekcji Vol. 10 pp. 47-56.
Paszotta Z. 2000b. Method Exterior Orientation of Aerial
Images by matching orthoimages. Wydawnictwo UWM,
Olsztyn , pp 6.
Paszotta, Z., Szumilo, M., 2006, Weryfikacja Numerycznego
Modelu Terenu.
Archiwum Fotogrametrii, Kartografii i
Teledetekcji Vol. 16 pp. 145-152.
Paszotta, Z., Levin, E., et al., 2010 Networed approach for rapid
geospatial imagery orthorectification. – in press
Peng, Z.,Tsou, M., 2003. Internet GIS, Distributed Geographic
Information services for the Internet and Wireless Networks.
John Wiley & Sons, New Jersey, pp. 120-123.
Petersen,
J.,
2009
ColdFusion
Article
“Benefits
of using the n-tiered approach for web applications”,
http://www.adobe.com/devnet/coldfusion/articles/ntier.html
Figure 8. Figure placement and numbering
Orthorectification was made on the server side but the grid
coordinate system was superimposed on the orthoimage by
means of applet on the client side. In both cases the image and
ground coordinates of point shown by the cursor are calculated
and printed in the header of the Internet browser window.
6. CONCLUSIONS
The process of creating orthophoto image together with DTM
from aerial photographs is complex and requiring the specialist
software on the digital photogrammetric station. However, it
turns out that this process with certain limitations can be
executed over the Internet. Even, if there is a simplified
solution, its advantages are great. This paper presented the
algorithm and web application for constructing DTM and
orthoimage online. The majority advantege of a such solution is
the ability to update and maintain it without distributing and
installing software on potentially thousands of client computers.